Total cost of ownership: SaaS vs custom software, properly compared

Quick answer: a fair SaaS-vs-custom comparison runs over five years, includes licences across all relevant tools, the integration layer wiring them together, internal staff time, and the migration cost when you eventually leave. Done that way, the breakeven for custom typically lands somewhere between fifty and two hundred staff, depending on the use case.

When clients ask us “is custom software cheaper than SaaS?”, the honest answer is it depends, but the comparison most people do is structurally wrong. Comparing “licence per month” to “build quote” misses 60% of what each costs to actually run.

Here’s the framework we use with clients evaluating their options.

Gartner’s definition of Total Cost of Ownership is the canonical reference for the framework. The application to software is straightforward; the data most teams skip is.

Why the simple comparison misleads

The default comparison is something like:

• SaaS: $50/user/month × 30 users = $1,500/month = $18,000/year
• Custom: $80,000 to build = 4.5 years to break even
• Therefore: stay on SaaS

This is the wrong maths because it ignores everything around the licence and everything around the build. A fair comparison includes both.

The full SaaS cost stack

Five-year SaaS total cost for a real business looks more like:

• Licence cost. The visible figure. Includes per-seat fees, tier upgrades over the period (most growing businesses are forced up at least one tier), and annual price increases (typically 5–10% per year on contract renewal).
• Adjacent platform cost. The tools that have to exist alongside the platform — landing-page builder, form tool, calendar, SMS provider, reporting / BI layer, dedicated email-sending service. These rarely fit inside the “main” platform pricing and add up to 20–50% on top of the headline licence.
• Integration cost. Wiring SaaS tools together — either through low-code platforms (Zapier, Make) at their own per-task pricing, or through engineering hours building proper integrations. For a stack of 8–12 tools, this is typically $500–$5,000/month plus build hours.
• Internal time. Staff time spent maintaining the platform — configuration, user management, training new starters, working around limitations, and (often) keeping a spreadsheet that fills the platform’s reporting gaps.
• Migration cost (deferred). The cost of leaving the platform when you eventually outgrow it. Discounted, this is often 6–12 months of the licence cost.

For a 50-staff business at $50/user/month on a marketing/CRM platform, the visible licence is about $30,000/year. The full stack — licences + adjacent tools + integration layer + internal time + amortised migration — tends to land between $80,000 and $150,000/year.

The full custom cost stack

Five-year custom-build cost looks like:

• Build cost. The one-off engineering investment. For a system replacing a meaningful portion of a SaaS platform’s functionality, typically $80,000–$300,000 depending on complexity.
• Hosting and infrastructure. Modest in most cases — serverless infrastructure for a 50-staff business is usually $200–$1,500/month all-in.
• Maintenance and ongoing engineering. Bug fixes, security patches, small feature additions, framework upgrades. Typically 10–25% of the build cost annually.
• Internal time. Staff time on the custom system — lower than on SaaS in our experience because the system was built around how the team actually works, but not zero.
• Replacement cost (deferred). No system lasts forever. Plan for a major rebuild or significant modernisation around year 5–7.

A typical $150,000 build for a 50-staff business amortises over five years at roughly $35,000/year in build, plus $10,000–$30,000/year in maintenance, plus $5,000–$15,000/year in hosting — total $50,000–$80,000/year. That’s often less than the equivalent SaaS stack at the same scale.

A worked example

A client we worked with [INSERT: industry, year, staff size]: their SaaS stack covering CRM, marketing automation, calendar/scheduling, customer messaging, and reporting was costing them about $9,500/month in licences. Layered on:

• $1,800/month in low-code automation tools and integrations
• About 40 hours/month of an internal operations FTE working around limitations
• A reporting tool for the gaps the CRM didn’t cover
• Two integrations that broke at every quarter-end

Total annual run rate: roughly $180,000 once everything was counted.

A custom replacement covering 80% of the functionality (the 20% they weren’t actually using) came in at $130,000 to build, $25,000/year to maintain, and $400/month to host. Amortised over five years: roughly $61,000/year. Roughly 3× cheaper than the SaaS stack while doing the work the team actually needed it to do.

This isn’t typical for a 10-staff business. It’s typical for a business at the scale where SaaS becomes a tax.

When the maths flips

The crossover point depends on five variables:

1. Headcount. Per-seat pricing scales with people; custom doesn’t.
2. Process specificity. The more your operation differs from the SaaS template, the more time you lose working around it — and the more value a custom system creates.
3. Integration density. The more SaaS tools have to talk to each other, the larger the integration layer cost.
4. Build complexity. Some workflows are genuinely hard to build well. Some are surprisingly cheap. The decision turns on which yours is.
5. Time horizon. Custom builds front-load cost; SaaS spreads it. The longer you’ll use the system, the better custom looks.

Rough rules of thumb we’ve seen hold up:

• Under 20 staff: SaaS almost always wins. The build cost rarely amortises before you outgrow the requirements.
• 20–50 staff: depends heavily on the use case. Probably SaaS unless the process is genuinely your differentiator.
• 50–200 staff: worth doing the full TCO comparison properly. Custom often wins for the systems your business actually runs on; SaaS still wins for adjacent tools.
• 200+ staff: custom for anything strategic; SaaS for commodity functions (HR, payroll, basic CRM).

A simple TCO worksheet

For each candidate path, calculate:

Cost line	SaaS	Custom
Year 1 build / set-up	one-off licence + onboarding	full build cost
Annual licence / maintenance	recurring per-seat	10–25% of build cost
Adjacent tools	bolt-ons (estimate)	minimal
Integration layer	iPaaS / engineering	integrated by design
Internal staff time	high (working around)	moderate
Hosting	included	$200–$1,500/month
Migration (when leaving)	meaningful	meaningful

Sum each column over five years. Compare the totals.

If the SaaS total is materially higher, do the full build-vs-buy decision framework before committing to custom. The numbers favour custom only when the use case also favours custom — cost alone isn’t enough.

The trap most businesses fall into

The trap isn’t picking the wrong answer at year zero. It’s never re-running the comparison.

Most businesses pick SaaS at year zero (correctly, given their size) and then never review the decision as they grow. By year five they’re paying three times what they should be, with an integration layer that breaks twice a quarter, and have never had the conversation about whether the original choice still fits.

A reasonable habit: review the TCO every two years, or whenever you double headcount, or whenever you cross $5,000/month on any single platform. Not as a trigger to rip and replace — as a trigger to ask the question.

Common questions

Is custom software cheaper than SaaS? At small scale (under 20 staff), almost never. At medium-to-large scale (50+ staff with process complexity), often substantially cheaper over five years. The decision turns on headcount, process specificity, and time horizon — not on the headline numbers.

How do I calculate total cost of ownership for SaaS? Add: licences across all tools in the stack, adjacent platforms (landing pages, forms, SMS, reporting), integration layer cost (Zapier/Make/iPaaS or engineering), internal staff time spent on the platform, and a discounted estimate of migration cost. Compare that — not just licences — to a build quote.

At what business size does custom software start to make sense? Roughly 50 staff for systems your business genuinely runs on, much later (or never) for adjacent commodity tools. The crossover is highly use-case-specific; the framework matters more than the rule of thumb.

What’s included in the cost of building custom software? Build cost (engineering hours), discovery and design, hosting infrastructure, ongoing maintenance (typically 10–25% of build cost annually), modest internal time, and a planned modernisation around year 5–7. Detailed in our custom web apps service page.

How long does custom software take to pay back? For a build at the right scale, typically 18–36 months. Builds that don’t pay back are usually the ones that didn’t need to happen — SaaS would have served just as well. The framework above filters those out before you commit.

If you’re at the point of doing this comparison properly for your own business, start a project and we’ll run the TCO with you. Honest answer either way.